Summary: How does a group of strangers settle on a shared habit or workplace norm? A new study suggests it isn’t through imitation or complex calculation. Instead, researchers found that humans follow a two-stage process: an initial phase of “sampling” different behaviors followed by a decisive commitment once a specific mathematical threshold is met.
This threshold is governed by the Tolerance Principle, a rule originally used to explain how children learn grammar, now proven to dictate the tipping points of social conventions and cultural change.
Key Facts
- The Two-Stage Shift: People don’t just copy the last thing they saw. They act probabilistically (trying different things) until their experience hits a mental “good enough” threshold, at which point they commit to the rule and ignore future exceptions.
- The Tolerance Principle: This simple equation predicts exactly how much consistency a person needs to see to treat a pattern as a “rule” despite occasional conflicting evidence.
- Universal Mechanism: The same cognitive engine that helps a child learn that “-ed” makes a word past tense (while “tolerating” irregular verbs like went) is what adults use to adopt social conventions and workplace cultures.
- Tipping Points: The model can accurately estimate the size of a dissenting minority needed to overturn a dominant norm, providing a mathematical blueprint for how social change and “flips” in collective behavior occur.
Source: CUNY
A paper in Proceedings of the National Academy of Sciences (PNAS) offers a strikingly simple answer to a longstanding question: How do people learn and settle on shared social conventions, from everyday habits to workplace norms?
Researchers from the CUNY Graduate Center, the University of Pennsylvania, and Stanford University found that people do not primarily learn by copying others or by calculating the most likely choice. Instead, they follow a two-stage process — sampling behaviors at first, then committing once enough evidence accumulates.
The study shows that this shift is governed by a simple mathematical rule known as the Tolerance Principle, which predicts when people have seen enough regularity to treat a pattern as a rule despite some exceptions.
Originally developed to explain how children learn the grammar of their native language, the researchers found, it also predicts how adults adopt shared behaviors — and even how competing norms can overturn one another.
“People often assume that social learning is about imitation or careful optimization,” said Spencer Caplan, Linguistics professor at the CUNY Graduate Center and co-lead author of the study. “What we found is something more basic and more human: People explore different options, but once a pattern crosses the threshold of ‘good enough,’ they commit to it — and stick with it even when there’s some conflicting evidence.”
To study how conventions emerge, the researchers built computational models of different learning strategies and evaluated them against data from coordination experiments — including previously published studies as well as a new set of experiments the team conducted.
In these experiments, participants had to independently align on shared choices — such as agreeing on a name for an unfamiliar face — while interacting in social networks. Participants received small rewards for matching others’ responses, allowing researchers to track how decisions evolved over time.
Across multiple experiments, people consistently deviated from the two dominant theories of social learning. They did not simply copy the most recent behavior they observed, nor did they always choose the statistically optimal option. Instead, they behaved probabilistically at first, reflecting uncertainty, and then made a decisive shift once their accumulated experience crossed a mental threshold.
That threshold is precisely captured by the Tolerance Principle: a simple equation that helps explain when people decide they’ve seen enough consistency to treat something as a rule even if there are a few exceptions.
The model not only better matched how people actually learn but also outperformed competing approaches, including Bayesian models, in predicting human behavior in controlled experiments.
The findings suggest that a single cognitive mechanism may underlie how people learn across domains, from language to social norms. Just as children learn to say “walked” and “talked” but also know that “went” is the past tense of “go” — applying a general rule while tolerating exceptions — adults appear to adopt conventions in a similarly rule-based way once they reach a point of sufficient evidence.
The research also sheds light on how social change happens. Because the model predicts when people commit to a convention, it can also estimate how large a dissenting minority needs to be to overturn it, offering new insight into tipping points in collective behavior.
“These results give us a clearer picture of how norms spread, stabilize, and sometimes flip,” Caplan said. “That has real implications for everything from public health campaigns to organizational culture to how new ideas take hold in society.”
The authors note that future work will examine how these dynamics play out in more complex, real-world settings, where social conventions are shaped not just by coordination but also by factors like identity, status, and power.
Key Questions Answered:
A: Because humans are “samplers.” We don’t want to commit to a norm that might be a fluke. Your brain waits until the evidence passes the “Tolerance” threshold. Once you’ve seen the behavior enough times to satisfy the math, you “click” into the norm and will likely stick with it even if you see someone else do it differently later.
A: Exactly. Once a norm crosses the Tolerance threshold, people stop “sampling” and start “applying the rule.” At that point, they become resistant to conflicting evidence. To “flip” the culture, you need a minority large enough to push the majority’s experience back across that mathematical threshold.
A: It’s the same math. A child hears “walked,” “played,” and “talked.” Eventually, their brain decides “-ed” is the rule. Even when they hear an exception like “ran” or “went,” they tolerate those exceptions because the “rule” has already reached sufficient mathematical weight in their mind.
Editorial Notes:
- This article was edited by a Neuroscience News editor.
- Journal paper reviewed in full.
- Additional context added by our staff.
About this social neuroscience research news
Author: Shawn Rhea
Source: CUNY
Contact: Shawn Rhea – CUNY
Image: The image is credited to Neuroscience News
Original Research: Closed access.
“A Simple Threshold Captures the Social Learning of Conventions” by Douglas Guilbeault, Spencer Caplan, and Charles Yang. PNAS
DOI:10.1073/pnas.2508061123
Abstract
A Simple Threshold Captures the Social Learning of Conventions
A persistent puzzle throughout the cognitive and social sciences is how people manage to learn social conventions from the sparse and noisy behavioral data of diverse actors, without explicit instruction.
Here, we show that the dominant theories of social learning perform poorly at capturing how individuals learn conventions in coordination experiments that task them with matching their behaviors while interacting in social networks.
Across experiments, participants’ choices systematically deviate from both imitation and optimization. Instead, they follow a categorical, two-stage learning process: they behave probabilistically until they acquire enough information about each other to trigger a mental threshold and then their choices stabilize.
We effectively estimate this threshold using the tolerance principle (TP), a parameter-free equation developed to model how children learn rules in language. We show that threshold-based agents produce social learning that is more accurate than imitating and optimizing agents, while also providing a better model of how a critical mass of dissenters can overturn conventions.
The superior performance of our model holds when comparing against a variety of optimization approaches, including Bayesian inference. Furthermore, in a preregistered dyadic experiment requiring people to infer nonlinguistic behavioral patterns amid controlled levels of noise in observed signals, TP outperforms all other models at reproducing learning rates among human participants.
These findings offer compelling evidence that a simple, mathematical threshold underlies individual and social learning, from grammatical rules to behavioral conventions.
